Kaplan-Meier curves for the cumulative incidence of any arrhythmia according to daily coffee intake after adjusting for basic demographic characteristics (age, sex, and ethnicity), body mass index, educational level, other comorbid conditions (hypertension, diabetes, hyperlipidemia, coronary heart disease, congestive heart failure, valvular heart disease, cerebrovascular disease, peripheral artery disease, chronic kidney disease, and cancer), smoking habits, alcohol consumption, tea consumption, and physical activity.
The relative hazard of each incident arrhythmia for each increasing category of coffee consumption (0-1, 2-3, 4-5, or ≥6 cups per day) is shown. Hazard ratios (HRs) are adjusted for basic demographic characteristics (age, sex, and ethnicity), body mass index, educational level, other comorbid conditions (hypertension, diabetes, hyperlipidemia, coronary heart disease, congestive heart failure, valvular heart disease, cerebrovascular disease, peripheral artery disease, chronic kidney disease, and cancer), smoking habits, alcohol consumption, tea consumption, and physical activity. Error bars indicate 95% CIs.
Caffeine is estimated to be the most widely used legal psychoactive drug in the world. About 85% of US adults consume caffeine daily, averaging about 135 mg per day, or about a 12 fl oz cup of coffee. This video explores what caffeine is and some ways it affects our bodies.
Click the Related Article link to read the associated Patient Page.
eMethods. Supplemental Methods
eTable 1. Coffee Intake Behavior per Coffee Type (Total N=502,543)
eTable 2. ICD-9 and ICD-10 Codes
eTable 3. Polygenic Score Quartile and Coffee Intake
eTable 4. CYP1A2 rs762551 and Coffee Intake
eTable 5. Interaction Analysis on the Relationship Between Coffee Consumption and Incident Arrhythmia by Genotype
eTable6. Polygenic Score of Caffeine Metabolism and Risk of Incident Arrhythmia
eTable 7. Caffeine Metabolism Determined by the CYP1A2 Variant and Risk of Incident Arrhythmia
eTable 8. Mendelian Randomization Analysis
eTable 9. Relationship Between Coffee Consumption or Genetic Variants for Caffeine Metabolism and the Risk of Arrhythmias Among Participants With Prevalent Arrhythmias (N=1,346)
eTable 10. Relationship Between Coffee Consumption of Genetic Variants for Caffeine Metabolism and Incident Arrhythmia in the Youngest Quartile of the Cohort (Mean Age 41± 0.75)
eTable 11. Relation Between Coffee Consumption or Genetic Variant for Caffeine Metabolism and the Risk of Incident Arrhythmia Among Males (N=182,715) and Females (N=201,255)
eFigure. Study Population
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Kim E, Hoffmann TJ, Nah G, Vittinghoff E, Delling F, Marcus GM. Coffee Consumption and Incident Tachyarrhythmias: Reported Behavior, Mendelian Randomization, and Their Interactions. JAMA Intern Med. 2021;181(9):1185–1193. doi:10.1001/jamainternmed.2021.3616
Is moderate, habitual coffee intake associated with the risk of arrhythmia, and is that association modified by genetic variants that affect caffeine metabolism?
In this large, prospective, population-based community cohort study of more than 300 000 participants, each additional daily cup of coffee was associated with a 3% reduced risk of developing an arrhythmia; these associations were not significantly modified by genetic variants that affect caffeine metabolism. A mendelian randomization study leveraging a polygenic score to capture inherited caffeine metabolism patterns did not reveal evidence that caffeine consumption increases the risk of incident arrhythmias.
Neither habitual coffee consumption nor genetically mediated differences in caffeine metabolism was associated with a heightened risk of cardiac arrhythmias.
The notion that caffeine increases the risk of cardiac arrhythmias is common. However, evidence that the consumption of caffeinated products increases the risk of arrhythmias remains poorly substantiated.
To assess the association between consumption of common caffeinated products and the risk of arrhythmias.
Design, Setting, and Participants
This prospective cohort study analyzed longitudinal data from the UK Biobank between January 1, 2006, and December 31, 2018. After exclusion criteria were applied, 386 258 individuals were available for analyses.
Daily coffee intake and genetic polymorphisms that affect caffeine metabolism.
Main Outcomes and Measures
Any cardiac arrhythmia, including atrial fibrillation or flutter, supraventricular tachycardia, ventricular tachycardia, premature atrial complexes, and premature ventricular complexes.
A total of 386 258 individuals (mean [SD] age, 56  years; 52.3% female) were assessed. During a mean (SD) follow-up of 4.5 (3.1) years, 16 979 participants developed an incident arrhythmia. After adjustment for demographic characteristics, comorbid conditions, and lifestyle habits, each additional cup of habitual coffee consumed was associated with a 3% lower risk of incident arrhythmia (hazard ratio [HR], 0.97; 95% CI, 0.96-0.98; P < .001). In analyses of each arrhythmia alone, statistically significant associations exhibiting a similar magnitude were observed for atrial fibrillation and/or flutter (HR, 0.97; 95% CI, 0.96-0.98; P < .001) and supraventricular tachycardia (HR, 0.96; 95% CI, 0.94-0.99; P = .002). Two distinct interaction analyses, one using a caffeine metabolism–related polygenic score of 7 genetic polymorphisms and another restricted to CYP1A2 rs762551 alone, did not reveal any evidence of effect modification. A mendelian randomization study that used these same genetic variants revealed no significant association between underlying propensities to differing caffeine metabolism and the risk of incident arrhythmia.
Conclusions and Relevance
In this prospective cohort study, greater amounts of habitual coffee consumption were associated with a lower risk of arrhythmia, with no evidence that genetically mediated caffeine metabolism affected that association. Mendelian randomization failed to provide evidence that caffeine consumption was associated with arrhythmias.
Coffee is 1 of the most widely consumed beverages worldwide.1 Although professional society guidelines that suggest avoiding caffeinated products to diminish the risk for arrhythmia2,3 have relied on assumed mechanisms and a small observational study4 from 1980, more recent investigations5,6 have consistently not demonstrated an increased risk of tachyarrhythmia among coffee consumers.
Of importance, coffee consumption may have multiple beneficial properties, often attributed to antioxidant and anti-inflammatory effects,7,8 and is associated with reduced risks of cancer,9 diabetes,10 Parkinson disease,11 and overall mortality.12 Indeed, the benefits appear to be most pronounced when caffeinated coffee is consumed.13 When coffee avoidance is recommended,14 we may withhold a beneficial substance that improves quality of life and longevity.12,15 We investigated the association of coffee intake with the risk of tachyarrhythmias in a large, population-based cohort, using participant self-report, mendelian randomization, and an analysis of related interactions to elucidate these associations.
The UK Biobank study was approved by the National Information Governance Board for Health and Social Care and the National Health Service North West Multicenter Research Ethics Committee. Written consent was obtained from all participants.16 The present analysis received certification from the University of California, San Francisco institutional review board to perform analyses of this deidentified data set. Race and/or ethnicity was self-reported at baseline surveys and obtained for the generalizability and applicability of the study results. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.17
We used the UK Biobank, a prospective study of participants in the UK National Health Services 40 to 69 years of age who resided within 40 km of 22 assessment centers. The UK Biobank study has been described previously.16 In brief, 502 543 participants recruited between January 1, 2006, and December 31, 2010, completed questionnaires, underwent physical examinations, and provided biological samples. We excluded participants who were pregnant on study entry (because reported coffee consumption was less likely to reflect long-term habits), those who subsequently withdrew from the study, and those with missing data.
Information regarding coffee consumption was obtained from participant questionnaires using a touch screen (eTable 1 in the Supplement). We grouped the participants into 8 categories corresponding to daily coffee intake: 0, less than 1, 1, 2, 3, 4, 5, and 6 or more cups daily. We excluded those who answered “do not know” or “prefer not to answer.”
Quiz Ref IDThe primary outcome of interest was incident tachyarrhythmia, ascertained between January 1, 2006, and December 31, 2018, using International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes available from inpatient and outpatient records (eTable 2 in the Supplement). The presence of the first of the following diagnoses was used to define the presence of an incident tachyarrhythmia: atrial fibrillation or atrial flutter (AF), supraventricular tachycardia, ventricular tachycardia, premature atrial complexes, and premature ventricular complexes. Participants with preexisting diagnoses of any these arrhythmias based on the same ICD-9 or ICD-10 codes between January 1, 2000, and December 31, 2005, were excluded from the primary analyses. Secondary analyses examining the incidence of each tachyarrhythmia alone, excluding those with a prevalent diagnosis of that particular tachyarrhythmia, were also conducted.
Race and ethnicity were self-identified and ascertained from the baseline survey. Educational level was categorized as middle school graduate, high school graduate, college degree, other professional degrees, and none of the above. Body mass index was derived from height and weight measurements obtained during the initial assessment and calculated as weight in kilograms divided by height in meters squared. Physical activity was categorized by tertiles from response to the question on the “number of days per week of moderate physical activity greater than 10 minutes.” Hypertension, diabetes, hyperlipidemia, coronary heart disease, congestive heart failure, valvular heart disease, cerebrovascular disease, peripheral artery disease, chronic kidney disease, and cancer were determined by ICD-9 and ICD-10 codes from inpatient and/or outpatient visits between January 1, 2000, and December 31, 2005.
DNA was purified from biological samples obtained during the initial assessment, such as blood, urine, and saliva. The initial 49 950 participants were genotyped using the UK BiLEVE Axiom array (Affymetrix, later Thermo Fisher Scientific), and the remaining 438 427 participants were genotyped using the closely related UK Biobank Axiom array (Thermo Fisher Scientific). These 2 arrays share 95% common content covering approximately 800 000 single-nucleotide variations (SNVs) (formerly single-nucleotide polymorphisms) and insertion-deletion markers. Quality control and imputation were performed centrally by the Wellcome Trust Centre for Human Genetics.18 More than 95% of caffeine is metabolized by CYP1A2, and the variability of its enzymatic activity is strongly associated with the variability of caffeine metabolism.19-22 CYP1A2 rs762551 was directly genotyped with a call rate of 99.8% and a minor allele frequency of 0.27, and its distribution follows Hardy-Weinberg equilibrium (P = .75).23 Previous population-based genome-wide association studies for caffeine metabolism revealed 8 loci associated with self-reported coffee and caffeine intake near the aryl hydrocarbon receptor and CYP1A2.24,25 Among these, rs6968554 was in a moderate linkage disequilibrium (R2 = 0.95). Therefore, we constructed a caffeine metabolism polygenic score by combining 7 genome-wide significant variants, including rs4410790, rs10275488, rs2892838, rs12909047, rs35107470, rs2470893, and rs2472297. The weighted genetic scores were derived by summing the number of alleles multiplied by their β coefficients (effect size for plasma paraxanthine to caffeine ratio) obtained from a previously reported genome-wide association study meta-analysis for caffeine metabolites.20 A higher genetic score represents slower caffeine metabolism. CYP1A2 rs762551 was not included in the genetic score because it was in a moderate linkage disequilibrium with other CYP1A2 variants (R2 = 0.05-0.29).23 However, it was analyzed independently because this variant was studied extensively for the association of caffeine with disease outcomes in other studies.23,26,27Quiz Ref ID To validate these SNVs as genetic markers of coffee consumption, we tested the hypothesis that faster metabolizers consumed more coffee by comparing self-reported coffee consumption to the polygenic score. The mendelian randomization analysis was performed only among those who self-identified as White British and had similar genetic ancestry based on a principal component analysis (UK Biobank data field 22006) and excluding related individuals so that no 2 individuals were a third-degree relative or closer.
Normally distributed continuous variables are described as means (SDs) and compared using unpaired, 2-tailed t tests, and continuous variables without a normal distribution are described as medians (interquartile ranges [IQRs]). Categorical variables were compared using χ2 tests. Kaplan-Meier curves were constructed to illustrate adjusted cumulative incidence rates. Unadjusted linear regression models were used to examine the association between coffee intake and the number of alleles among individual genetic variants and the polygenic score, respectively. Multivariate Cox proportional hazards regression models were used to assess factors associated with incident events, adjusting for covariates that were likely to confound or mediate examined associations based on the previous literature and biological plausibility. Genetic analyses were adjusted for the top 15 principal components and the specific genotyping array used, as recommended.28 The main mendelian randomization analysis was performed using the inverse-variance weighted method in a fixed-effects model. We also performed sensitivity analyses to assess for pleiotropy using simple median, weighted median, mendelian randomization Egger, and mendelian randomization residual sum and outlier methods.28,29 Analyses were executed on all variants as well as 2 subsets separately (chromosome 7 and chromosome 1) and were performed separately using coffee coefficient estimates from our UK Biobank sample as well as an independent sample.20,23 A secondary analysis was performed without excluding related individuals and adjusted for the same covariates. The potential gene-coffee interaction was evaluated by assessing for effect modification using multivariate Cox proportional hazards regression time-dependent models for both the polygenic score and the genetic variant CYP1A2 (rs762551), including an interaction term between self-reported coffee consumed and the respective genetic variant covariate. Proportional hazards assumptions were checked with log-log survival plots rather than formal statistical tests for the association of time and Schoenfeld residuals, which are commonly oversensitive in very large data sets, signaling no violations of substantive importance.
The following sensitivity analyses were performed: analyses were performed again after excluding those who consumed decaffeinated coffee (to provide a comparison across varying amounts of caffeinated coffee, including no coffee), and all analyses were performed again after constraining outcome ascertainment to inpatient records alone and separately using outpatient records alone. We repeated the main analyses comparing those with and without prevalent arrhythmias using logistic regression. In addition, the analyses were performed again after restricting the cohort to the youngest quartile, and interaction testing to evaluate age as an effect modifier of the coffee-arrhythmia association was conducted. Analyses were performed again after stratifying by biological sex, and interaction testing by sex was performed. We performed multivariate Cox proportional hazards regression analysis as a positive control to assess smoking status as a factor associated with incident arrhythmia based on its established association with arrhythmia risk.30
A 2-tailed P < .05 was considered to be statistically significant. Statistical analyses were performed using SAS software, version 9.4 (SAS Institute Inc); Stata software, version 14.2 (StataCorp LLC); and R software, version 3.5.131 with R package Mendelian Randomization, version 0.3.0.32
Quiz Ref IDAmong 502 543 UK Biobank participants, a median of 2 cups (IQR, 1-4 cups) of coffee were consumed per day, and 111 218 participants (22.1%) did not drink coffee. Among coffee drinkers, 310 061 (61.7%) were caffeinated and 74 371 (14.8%) were noncaffeinated drinkers (eTable 1 in the Supplement). After exclusion criteria were applied, 386 258 participants (mean [SD] age, 56  years; 52.3% female) were eligible for the current analyses (eFigure in the Supplement). A total of 208 810 participants had only inpatient records available, and the remaining population had both inpatient and outpatient records available for analyses. Baseline characteristics of the cohort are given in Table 1. Those who consumed more than the daily median amount of coffee were more likely to be older, White, and male, have peripheral artery disease, have cancer, be a current smoker, or drink alcohol and less likely to have hypertension, diabetes, and chronic kidney disease and to drink tea.
During a mean (SD) of 4.5 (3.1) years of follow-up, 16 369 incident arrhythmias occurred: 12 811 AFs, 1920 supraventricular tachycardias, 909 ventricular tachycardias, 97 premature atrial complexes,632 premature ventricular complexes, and 610 unspecified arrhythmias. Quiz Ref IDAfter adjustment for basic age, sex, race, ethnicity, hypertension, diabetes, hyperlipidemia, coronary heart disease, congestive heart failure, valvular heart disease, cerebrovascular disease, peripheral artery disease, chronic kidney disease, cancer, educational level, smoking, alcohol and tea consumption, and physical activity, each additional cup of coffee intake was associated with a 3% lower risk of incident arrhythmia (hazard ratio [HR], 0.97; CI, 0.96-0.98; P < .001) (Figure 1 and Figure 2). In analyses of each arrhythmia alone, statistically significant associations exhibiting a similar magnitude were observed for AF (HR, 0.97; 95% CI, 0.96-0.98; P < .001) and supraventricular tachycardia (HR, 0.96; 95% CI, 0.94-0.99; P = .002) (Figure 2). Although the point estimates were similar for ventricular tachycardia and premature atrial complexes, those associations did not achieve statistical significance. In the positive control analysis using the same outcome ascertainment for arrhythmias and adjusting for the same covariates as were used in the coffee analyses, smokers experienced a significantly higher risk of incident arrhythmia (HR, 1.09; 95% CI, 1.03-1.15; P = .002).
Consistent with previous reports,20,23,26,33 those with genetic variants associated with slower caffeine metabolism consumed less coffee (eTables 3 and 4 in the Supplement). In models to assess possible modification by polygenic score or the CYP1A2 rS762551 variant on the association between coffee intake and arrhythmia risk, we found no meaningful differences in the point estimates among those with different genetic variants or any statistically significant interactions (eTable 5 in the Supplement).
In genetic analysis, higher weighted polygenic score (slower caffeine metabolism, separately associated with lower coffee consumption) did not reveal any significant association with arrhythmia risk (Table 2). Adjustment for the same covariates listed above without excluding related individuals did not alter this finding (eTable 6 in the Supplement). Analyses of the association between the CYP1A2 variant alone and incident arrhythmias revealed similar findings in both pruned and multivariable adjusted analyses (Table 3; eTable 7 in the Supplement). No meaningful differences in these results were observed in any of the mendelian randomization sensitivity analyses, with no evidence of directional pleiotropy (eTable 8 in the Supplement).
Additional sensitivity analyses examining prevalent arrhythmias similarly found a statistically significant 7% decreased odds (95% CI, 0.90-0.96; P < .001) of prevalent arrhythmia per each additional daily cup of coffee, and, as in the incident analyses, neither the polygenic score nor differences in the CYP1A2 variant revealed evidence of heightened risks of prevalent arrhythmias (eTable 9 in the Supplement). Results remained consistent whether restricting to the youngest quartile of the cohort, women only, or men only. No interactions by age or sex were observed (eTables 10 and 11 in the Supplement).
In this large, prospective, population-based cohort study, consuming more habitual coffee was associated with a lower risk of developing an arrhythmia. There was no evidence that the association between coffee intake and arrhythmia risk was affected by genetic variants associated with caffeine metabolism. The mendelian randomization analyses failed to provide evidence that caffeine consumption leads to a greater risk of arrhythmias.
Coffee serves as the primary source of caffeine for most individuals,34 and it has a reputation for causing or exacerbating arrhythmias.2,3 Caffeine as a proarrhythmic drug has biological plausibility given factors associated with the increase in serum catecholamine levels35,36 and the calcium release from the sarcoplasmic reticulum in a fashion that may lead to delayed afterdepolarizations.36 Early, small observational studies4,37 supported the concept that caffeine was associated with a higher arrhythmia risk. The Physician’s Health Study, limited to only male physicians, then suggested a U-shaped association between coffee consumption and arrhythmia risk, with the lowest risk of AF among those who consumed 1 to 3 cups daily.6 However, another study5 failed to replicate similar findings. A meta-analysis38 of 7 observational studies of 115 993 participants found that caffeine exposure based on coffee consumption did not affect AF risk. In a more recent Danish study,39 coffee consumption in general (compared with no consumption) was inversely associated with AF incidence. Furthermore, possible health benefits of caffeine and coffee in particular have emerged, such as reduction in cancer, diabetes, cardiovascular disease, and overall mortality.9-12 A review15 of 201 meta-analyses found that moderate consumption of coffee is likely more beneficial than harmful to health. However, the predominant perception among the public and health care professionals is that caffeine commonly triggers arrhythmias.4,14,40
Most research exploring the associations between coffee and arrhythmias has relied on self-reported coffee consumption. In addition, because most of these studies are observational rather than randomized trials, decisions to drink coffee and the amounts regularly consumed are likely associated with other variables that may confound observed associations. Mendelian randomization is a tool that may help mitigate these limitations: by identifying alleles in common genetic variants that are associated with the factor studied (in this case, coffee consumption) without inherent associations with the outcome (in this case, tachyarrhythmias), the random properties of meiosis in assigning genotype can be used to infer causality.41,42 Because other behaviors or circumstances during life cannot affect the genotype, confounding by such factors should not be operative. Such methods have been used in alcohol research, leveraging the observation that those who metabolize alcohol more quickly tend to drink more alcohol,43,44 and the technique has recently been applied to understanding caffeine metabolism.26,45
The findings of this study failed to provide any evidence that coffee consumption heightens the risk of developing arrhythmias. Instead, on the basis of self-report of coffee consumption, the results suggest that coffee intake may in fact reduce the risk of arrhythmias, with the strongest evidence pertaining to developing atrial arrhythmias and supraventricular tachycardia. These findings are consistent with previous studies6,38,46,47 that found no significant association with or a reduced risk of AF. Several possible mechanisms may be involved in the association between antiarrhythmic factors and caffeine. For example, AF is more likely to occur in the setting of a shorter atrial effective refractory period, whereas caffeine appears to prolong left atrial effective refractory periods.48 Caffeine is known to block adenosine receptors, high doses of adenosine are known to trigger AF, and reduced AF attributed to inhibiting adenosine receptors has been observed in animal models.49,50 Coffee’s antioxidant and anti-inflammatory properties may also play a role1,7,8,51 given that inflammation can be a sufficient substrate for cardiac arrhythmias via multiple mechanistic pathways.52,53 Finally, even the catecholaminergic properties of caffeine could be protective against some arrhythmias, such as some premature ventricular complexes,54 as well as some arrhythmias triggered by enhanced vagal tone.55
It is also possible that the reduced risk of arrhythmia among those who consumed more coffee in the present study occurred because of underascertainment of prevalent arrhythmias, particularly if this underascertainment led to abstention from caffeine among those with previous arrhythmias before study enrollment, leading to misclassification of prevalent cases as incident ones. It would appear unlikely, however, that such misclassification would explain a graded reduction in arrhythmia risk with more habitual coffee consumption among large numbers of coffee drinkers. In addition, this limitation cannot explain the mendelian randomization findings because these findings do not rely on self-reported coffee consumption but instead on inherited and static (within an individual) genetic polymorphisms.
It is plausible that some individuals experience coffee or caffeine-induced arrhythmias, whereas others do not. This study was unable to identify any modification of the association between coffee consumption and arrhythmia risk by different genetic determinants of coffee metabolism. However, there may be other genes, such as those related to arrhythmia risk, that may yet modify this association.
Mendelian randomization evaluating the association between genetic polymorphisms that affect caffeine metabolism and incident arrhythmia revealed neutral, nonsignificant associations, again supporting the absence of an association between caffeine consumption and arrhythmia risk. This approach was validated by our observation that those participants expected to metabolize caffeine more quickly did indeed consume more coffee. Multiple sensitivity analyses and multivariable adjustment of the analyses for polygenic score or CYP1A2 variant did not alter the association. Although it may be a true phenomenon, the results of mendelian randomization studies warrant a cautious interpretation. Mendelian randomization is considered a powerful tool to infer causality from nature’s randomization, but it is not completely protected from bias and confounders.42 For example, CYP1A2 activity is strongly affected by smoking status, sex, and ethnicity.21 However, confounding could not have resulted in the genetic variant, and it is possible that attenuation of a protective effect against arrhythmias may have been caused by adjusting for mediators that are actually along the causal pathway or associated with collider bias.45 In addition, the polygenic score explains only a relatively small proportion of the variation in coffee consumption.33 Finally, the use of caffeine metabolism variants as proxies for caffeine intake has inherent limitations because of opposing effects on coffee intake vs resultant circulating caffeine levels, including biological exposure to caffeine.20,25,34,45 In the end, however, this study found no consistent evidence from these genetic analyses that a greater propensity to consume caffeinated beverages increased the risk of arrhythmia.
Our study has multiple strengths. The study was a community-based prospective cohort study with an unprecedented sample size to investigate the association between coffee intake and incident arrhythmias, including the first such analyses of genetic interactions or a related mendelian randomization analysis. The data were carefully collected for potential confounders and mediators, including demographic characteristics, health, and lifestyle factors.
It is important to acknowledge several limitations of the study. Similar to nearly every epidemiologic study on the subject, coffee intake data were obtained from self-report. However, that information was obtained before the subsequent development of the outcomes of interest, so recall bias could not have been operative. The mendelian randomization analyses, despite the limitations already described, should also help to mitigate that limitation. The detailed information regarding the type of coffee, such as espresso or not, or other caffeinated products was not available. Quiz Ref IDThe analyses were performed with an assumption that the coffee consumption reported at baseline was sufficiently indicative of consumption that persisted during the study. Although this assumption is a limitation, previous studies56,57 have demonstrated consistent answers regarding coffee consumption over time, suggesting that a baseline assessment accurately captures long-term consumption patterns. Multiple studies56,58 evaluating different types of nutrient questionnaires revealed a high degree of reproducibility of nutrient intake over time. The effect of coffee intake on arrhythmia risk was assumed to be constant over time, which is challenging to prove. Although the risk of arrhythmias associated with caffeine is generally considered to be immediate (rather than a long-term event that only manifests over many years),54 the mean follow-up of just more than 4 years does not allow comments on the longer-term association. A substantial number of participants needed to be excluded because of missing data. Although constraining our study cohort to those with complete data available should help mitigate against threats to internal validity, these exclusions could limit generalizability. This study relied on ICD-9 and ICD-10 codes to ascertain the incident arrhythmia outcomes. Although these codes yield high sensitivities and specificities when compared with adjudication by detailed medical record review,59,60 they likely present a lack of sensitivity for certain arrhythmias, particularly premature atrial complexes (which would potentially require continuous electrocardiographic monitoring in all participants to achieve maximum accuracy). This lack of sensitivity may explain why significant differences in more common and more clinically evident arrhythmias, such as AF and supraventricular tachycardia, were observed. Of note, this limitation would be expected to reduce power to detect an association (therefore we cannot exclude a false-negative result in regard to less common arrhythmias) but should not have led to type I errors or spurious false-positive results. Although exhaustive adjustment was performed in the multivariable analyses, residual or unmeasured confounding cannot be excluded. Although mendelian randomization is purported to provide a methodologic approach wherein causality may be inferred from observational data, such conclusions should be cautioned against. Despite the sensitivity analyses, there may yet be pleiotropic effects, important variants in linkage disequilibrium with the SNVs that were studied, or insufficient power that explains the observations. Ultimately, only a randomized clinical trial can definitively demonstrate clear effects of coffee or caffeine consumption.
In this prospective cohort study, increasing amounts of habitual coffee intake were associated with a lower risk of arrhythmia, particularly for AF and supraventricular tachycardia, with no evidence that genetically determined differences in caffeine metabolism modified these associations. Mendelian randomization similarly failed to reveal any evidence that expected increases in coffee consumption would be associated with a higher risk of arrhythmia. These data suggest that common prohibitions against caffeine to reduce arrhythmia risk are likely unwarranted.
Accepted for Publication: May 23, 2021.
Published Online: July 19, 2021. doi:10.1001/jamainternmed.2021.3616
Corresponding Author: Gregory M. Marcus, MD, MAS, Division of Cardiology, University of California, San Francisco, 400 Parnassus Ave, San Francisco, CA 94122 (email@example.com).
Author Contributions: Drs Kim and Marcus had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Kim, Marcus.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Kim, Marcus.
Critical revision of the manuscript for important intellectual content: Kim, Hoffmann, Nah, Vittinghoff, Delling.
Statistical analysis: Kim, Hoffmann, Nah, Vittinghoff.
Obtained funding: Marcus.
Administrative, technical, or material support: Delling, Marcus.
Conflict of Interest Disclosures: Dr Vittinghoff reported receiving salary support from the National Institutes of Health during the conduct of the study. Dr Marcus reported receiving grants from Baylis, Medtronic, and Eight Sleep outside the submitted work and reported being a consultant for Johnson & Johnson and InCarda and holding equity in InCarda. No other disclosures were reported.
Funding/Support: This research was conducted using the UK Biobank resource. The UK Biobank was established by the Wellcome Trust, the Medical Research Council, the UK Department of Health, and the Scottish Government. The UK Biobank has also received funding from the Welsh Assembly Government, the British Heart Foundation, and Diabetes United Kingdom, Northwest Regional Development Agency, Scottish Government.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Information: Dr Kim is currently in cardiac electrophysiology fellowship training at the University of California, San Francisco, San Francisco.